add keyword table s3 storage support

api/.env.example

@@ -137,3 +137,4 @@ SSRF_PROXY_HTTP_URL=
 SSRF_PROXY_HTTPS_URL=
 
 BATCH_UPLOAD_LIMIT=10
+KEYWORD_DATA_SOURCE_TYPE=database

api/config.py

@@ -62,6 +62,7 @@
     'KEYWORD_STORE': 'jieba',
     'BATCH_UPLOAD_LIMIT': 20,
     'TOOL_ICON_CACHE_MAX_AGE': 3600,
+    'KEYWORD_DATA_SOURCE_TYPE': 'database',
 }
 
 
@@ -303,6 +304,7 @@ def __init__(self):
         self.API_COMPRESSION_ENABLED = get_bool_env('API_COMPRESSION_ENABLED')
         self.TOOL_ICON_CACHE_MAX_AGE = get_env('TOOL_ICON_CACHE_MAX_AGE')
 
+        self.KEYWORD_DATA_SOURCE_TYPE = get_env('KEYWORD_DATA_SOURCE_TYPE')
 
 class CloudEditionConfig(Config):
 

api/core/rag/datasource/keyword/jieba/jieba.py

@@ -2,13 +2,15 @@
 from collections import defaultdict
 from typing import Any, Optional
 
+from flask import current_app
 from pydantic import BaseModel
 
 from core.rag.datasource.keyword.jieba.jieba_keyword_table_handler import JiebaKeywordTableHandler
 from core.rag.datasource.keyword.keyword_base import BaseKeyword
 from core.rag.models.document import Document
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
+from extensions.ext_storage import storage
 from models.dataset import Dataset, DatasetKeywordTable, DocumentSegment
 
 
@@ -108,6 +110,9 @@ def delete(self) -> None:
         if dataset_keyword_table:
             db.session.delete(dataset_keyword_table)
             db.session.commit()
+            if dataset_keyword_table.data_source_type != 'database':
+                file_key = 'keyword_files/' + self.dataset.tenant_id + '/' + self.dataset.id + '.txt'
+                storage.delete(file_key)
 
     def _save_dataset_keyword_table(self, keyword_table):
         keyword_table_dict = {
@@ -118,28 +123,41 @@ def _save_dataset_keyword_table(self, keyword_table):
                 "table": keyword_table
             }
         }
-        self.dataset.dataset_keyword_table.keyword_table = json.dumps(keyword_table_dict, cls=SetEncoder)
-        db.session.commit()
+
+        keyword_data_source_type = current_app.config['KEYWORD_DATA_SOURCE_TYPE']
+        if keyword_data_source_type == 'database':
+            self.dataset.dataset_keyword_table.keyword_table = json.dumps(keyword_table_dict, cls=SetEncoder)
+            db.session.commit()
+        else:
+            file_key = 'keyword_files/' + self.dataset.tenant_id + '/' + self.dataset.id + '.txt'
+            if storage.exists(file_key):
+                storage.delete(file_key)
+            storage.save(file_key, json.dumps(keyword_table_dict, cls=SetEncoder).encode('utf-8'))
 
     def _get_dataset_keyword_table(self) -> Optional[dict]:
         lock_name = 'keyword_indexing_lock_{}'.format(self.dataset.id)
         with redis_client.lock(lock_name, timeout=20):
             dataset_keyword_table = self.dataset.dataset_keyword_table
             if dataset_keyword_table:
-                if dataset_keyword_table.keyword_table_dict:
-                    return dataset_keyword_table.keyword_table_dict['__data__']['table']
+                keyword_table_dict = dataset_keyword_table.keyword_table_dict
+                if keyword_table_dict:
+                    return keyword_table_dict['__data__']['table']
             else:
+                keyword_data_source_type = current_app.config['KEYWORD_DATA_SOURCE_TYPE']
                 dataset_keyword_table = DatasetKeywordTable(
                     dataset_id=self.dataset.id,
-                    keyword_table=json.dumps({
+                    keyword_table='',
+                    data_source_type=keyword_data_source_type,
+                )
+                if keyword_data_source_type == 'database':
+                    dataset_keyword_table.keyword_table = json.dumps({
                         '__type__': 'keyword_table',
                         '__data__': {
                             "index_id": self.dataset.id,
                             "summary": None,
                             "table": {}
                         }
                     }, cls=SetEncoder)
-                )
                 db.session.add(dataset_keyword_table)
                 db.session.commit()
 

api/extensions/ext_storage.py

@@ -172,6 +172,20 @@ def exists(self, filename):
 
             return os.path.exists(filename)
 
+    def delete(self, filename):
+        if self.storage_type == 's3':
+            self.client.delete_object(Bucket=self.bucket_name, Key=filename)
+        elif self.storage_type == 'azure-blob':
+            blob_container = self.client.get_container_client(container=self.bucket_name)
+            blob_container.delete_blob(filename)
+        else:
+            if not self.folder or self.folder.endswith('/'):
+                filename = self.folder + filename
+            else:
+                filename = self.folder + '/' + filename
+            if os.path.exists(filename):
+                os.remove(filename)
+
 
 storage = Storage()
 

api/migrations/versions/17b5ab037c40_add_keyworg_table_storage_type.py

@@ -0,0 +1,33 @@
+"""add-keyworg-table-storage-type
+
+Revision ID: 17b5ab037c40
+Revises: a8f9b3c45e4a
+Create Date: 2024-04-01 09:48:54.232201
+
+"""
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = '17b5ab037c40'
+down_revision = 'a8f9b3c45e4a'
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+
+    with op.batch_alter_table('dataset_keyword_tables', schema=None) as batch_op:
+        batch_op.add_column(sa.Column('data_source_type', sa.String(length=255), server_default=sa.text("'database'::character varying"), nullable=False))
+
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+
+    with op.batch_alter_table('dataset_keyword_tables', schema=None) as batch_op:
+        batch_op.drop_column('data_source_type')
+
+    # ### end Alembic commands ###

api/models/dataset.py

@@ -1,11 +1,13 @@
 import json
+import logging
 import pickle
 from json import JSONDecodeError
 
 from sqlalchemy import func
 from sqlalchemy.dialects.postgresql import JSONB, UUID
 
 from extensions.ext_database import db
+from extensions.ext_storage import storage
 from models.account import Account
 from models.model import App, UploadFile
 
@@ -441,6 +443,7 @@ class DatasetKeywordTable(db.Model):
     id = db.Column(UUID, primary_key=True, server_default=db.text('uuid_generate_v4()'))
     dataset_id = db.Column(UUID, nullable=False, unique=True)
     keyword_table = db.Column(db.Text, nullable=False)
+    data_source_type = db.Column(db.String(255), nullable=False, server_default=db.text("'database'::character varying"))
 
     @property
     def keyword_table_dict(self):
@@ -454,8 +457,24 @@ def object_hook(self, dct):
                         if isinstance(node_idxs, list):
                             dct[keyword] = set(node_idxs)
                 return dct
-
-        return json.loads(self.keyword_table, cls=SetDecoder) if self.keyword_table else None
+        # get dataset
+        dataset = Dataset.query.filter_by(
+            id=self.dataset_id
+        ).first()
+        if not dataset:
+            return None
+        if self.data_source_type == 'database':
+            return json.loads(self.keyword_table, cls=SetDecoder) if self.keyword_table else None
+        else:
+            file_key = 'keyword_files/' + dataset.tenant_id + '/' + self.dataset_id + '.txt'
+            try:
+                keyword_table_text = storage.load_once(file_key)
+                if keyword_table_text:
+                    return json.loads(keyword_table_text.decode('utf-8'), cls=SetDecoder)
+                return None
+            except Exception as e:
+                logging.exception(str(e))
+                return None
 
 
 class Embedding(db.Model):